A Lightweight Encoder–Decoder Image Automatic Colorization Framework with Global Palette Modulation and Luminance Consistency
Advances in Engineering Technology Research · 11.03.2026 · 16 (1) · 683
DOI: 10.56028/aetr.16.1.683.2026 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advances in Engineering Technology Research, online erschienen am 11.03.2026. Verfasst von Zhangyan Wang.
Abstract (Originalsprache)
Image colorization is an important research topic in the field of computer vision and has been widely applied in scenarios such as old photo restoration, surveillance video enhancement, medical image visualization, and industrial vision assistance. This paper proposes an end-to-end image colorization method based on deep convolutional neural networks (CNNs), aiming to automatically generate color images from grayscale inputs. Experiments are conducted on the CIFAR-10 dataset using a lightweight Encoder–Decoder network architecture. Two novel techniques, Global Palette Modulation (GPM) and Luminance Consistency Loss (Luminance Loss), are introduced, significantly improving the visual quality and structural fidelity of the colorization results. First, to capture global scene color information, the proposed GPM module modulates high-level semantic features at the network bottleneck, generating an adaptive color palette for each image and enhancing the model’s ability to express overall color style. Second, Luminance Consistency Loss is incorporated into the training process to ensure that the generated color image preserves the luminance structure of the original grayscale image, thereby retaining object contours and textures. Comparative experiments demonstrate that the proposed method with GPM and luminance consistency loss achieves significant improvements across multiple evaluation metrics, including MSE, MAE, PSNR, and SSIM. Experimental results show that the proposed approach not only produces high-quality colorization results but also maintains low computational complexity, making it suitable for rapid prototyping in educational and engineering applications.
Zitieren
APA
Wang, Z. (2026). A Lightweight Encoder–Decoder Image Automatic Colorization Framework with Global Palette Modulation and Luminance Consistency. Advances in Engineering Technology Research, 16(1), 683. https://doi.org/10.56028/aetr.16.1.683.2026
Vancouver
Wang Z. A Lightweight Encoder–Decoder Image Automatic Colorization Framework with Global Palette Modulation and Luminance Consistency. Advances in Engineering Technology Research. 2026;16(1):683. doi:10.56028/aetr.16.1.683.2026
BibTeX
@article{Wang2026,
author = {Wang, Zhangyan},
title = {A Lightweight Encoder–Decoder Image Automatic Colorization Framework with Global Palette Modulation and Luminance Consistency},
journal = {Advances in Engineering Technology Research},
year = {2026},
volume = {16},
number = {1},
pages = {683},
doi = {10.56028/aetr.16.1.683.2026}
}
EUVIMED